PDFs, the LHC and the Higgs

Transcription

1 PDFs, the LHC and the Higgs J. Huston Michigan State University MCTP Spring Symposium on Higgs Physics April 17, 2012

2 Some references and companion website at

3 Some references YR1 YR2

4 Understanding cross sections at the LHC We re all looking for Higgs/BSM physics at the LHC Before we publish Higgs/BSM discoveries from the early running of the LHC, we want to make sure that we measure/understand the relevant cross sections and this largely means understanding QCD at the LHC in final states involving vector bosons, jets, photons, heavy quarks 2010 was largely spent Rediscovering the Standard Model at the LHC my phrase by the way, so reference me if you use it 2011 was spent in extending the discovery reach beyond the Tevatron An important part of this understanding is related to our understanding of parton distribution functions and their uncertainties for LHC kinematics

5 The re-discovery Each of these cross sections is compared to either NLO or NNLO predictions using current PDFs Many of these cross sections are being fed back into the global PDF fits providing information in a new kinematic regime but only if the experiment provides the needed correlated systematic error info

6 Parton distribution functions and global fits Calculation of production cross sections at the LHC relies upon knowledge of PDF s in the relevant kinematic region PDF s are determined by global analyses of data from DIS, DY and jet production There are three major groups that provide semi-regular updates to global fits to parton distributions when new data/theory becomes available MRS->MRST2004->MSTW2008 CTEQ->->CTEQ6.6->CT09- >CT10->CT12 NNPDF->NNPDF2.0->NNPDF2.1 There are three other groups that also provide updated PDF fits, that are not quite as fully global HERAPDF ABKM (G)JR 1. BCDMS F 2 proton (339 data points) 2. BCDMS F 2 deuteron (251 data points) 3. NMC F 2 (201 data points) 4. NMC F 2d /F 2 p (123 data points) 5. F 2 (CDHSW) (85 data points) 6. F 3 (CDHSW) (96 data points) 7. CCFR F 2 (69 data points) 8. CCFR F 3 (86 data points) 9. H1 NC e - p (126 data points; reduced cross section) 10. H1 NC e - p (13 data points; high y analysis) 11. H1 NC e + p (115 data points; reduced cross section ) 12. H1 NC e + p (147 data points; reduced cross section; ) 13. ZEUS NC e - p (92 data points; ) 14. ZEUS NC e + p (227 data points; ) 15. ZEUS NC e + p (90 data points; ) 16. H1 F c 2 e + p (8 data points; ) 17. H1 Rσ c for ccbar e + p (10 data points; ) 18. H1 R b σ for bbbar e + p (10 data points; ) 19. ZEUS F c 2 e + p (18 data points; 1996/97) 20. ZEUS F C 2 e+p (27 data points; 1998/00) 21. H1 CC e - p (28 data points; ) 22. H1 CC e + p (25 data points; ) 23. H1 CC e + p (28 data points; ) 24. ZEUS CC e - p (26 data points; ) 25. ZEUS CC e + p (29 data points; ) 26. ZEUS CC e + p (30 data points; ) 27. NuTev neutrino dimuon cross section (38 data points) 28. NuTev anti-neutrino dimuon cross section (33 data points) 29. CCFR neutrino dimuon cross section (40 data points) 30. CCFR anti-neutrino cross section (38 data points) 31. E605 dimuon (199 data points) 32. E866 dimuon (13 data points) 33. Lepton asymmetry from CDF (11 data points) 34. CDF Run 1B jet cross section (33 data points) 35. D0 Run 1B jet cross section (90 data points) Some of the PDF groups use a Hessian approach towards the estimation of errors; others use a Monte Carlo approach. CTEQ6.6 data sets CT10 also includes Run 2 jets, HERA 1 combined data; CT12 uses LHC data in addition Some use a more limited parameterization for their PDFs.

7 α s The value of α s (m Z ) used in global fits can either be determined by the global fit (a la MSTW), or input as an external parameter (a la CTEQ-> world average) Global fits can not determine the value of α s (m Z ) very precisely The world average in 2009 was / In 2011, it is / This is a mixture of analyses and calculations at different orders, including lattice calculations perhaps α s (m Z ) at NNLO should be smaller than at NLO For the PDF4LHC report, we have used a +/ variation as a 90% CL error ( as a 1-sigma error) The error in α s can be added in quadrature with the PDF error, without approximation G. Watt arxiv:

8 PDF4LHC report We carried out an exercise to which all PDF groups were invited to participate A comparison of NLO predictions for benchmark cross sections at the LHC (7 TeV) using MCFM with prescribed input files Benchmarks included W/Z production/rapidity distributions ttbar production Higgs production through gg fusion masses of 120, 180 and 240 GeV PDFs used include CTEQ6.6, MSTW08, NNPDF2.0, HERAPDF1.0, ABKM09, GJR08 In some of comparisons, updates to above PDFs may also be shown

9 PDF luminosities As a first step, it s useful to define and to compare PDF luminosities from the different PDF groups

10 PDF luminosities The qq luminosities for the groups tend to have different behaviors at low mass and at high mass The reasons can often be understood NNPDF2.0 does not use a heavy quark flavor scheme; this suppresses the low x quark and anti-quark distributions (NNPDF2.1 does use such a scheme) HERAPDF uses the HERA combined Run 1 dataset that prefers a higher normalization The agreement tends to be much better in the W/Z region Plots by G. Watt arxiv:

11 PDFs are tending to get closer NNPDF2.1 has a GM-VFNS treatment (FONLL) ->increase in low x quarks CT10 includes Tevatron Run II jet data

12 PDFs Larger differences are observed for gg luminosities, especially at high mass critically depends on whether Tevatron inclusive jet data have been used or not Plots by G. Watt arxiv:

13 Uncertainties Uncertainties, at least among the global PDF groups, agree amazingly well for qqbar, especially given different approaches/assumptions A bit larger spread for gg Unless otherwise stated, all PDF uncertainties are at 68% CL some PDF groups produce uncertainties for both 68% and 90%CL for others, a scaling of is used (which works well)

14 Cross section comparisons Notice that the CTEQ and MSTW predictions for W/Z production are very close to each other Also, in general, there is very little dependence of the cross sections on the value of α s (m Z ) (as expected) And of course, the higher qq luminosities observed earlier lead to higher predictions for W/Z cross sections for HERAPDF

15 Cross section comparisons Larger gg differences and greater dependence on α s lead to larger differences in Higgs/tT cross section Note that there tends to be two groupings Plots by G. Watt arxiv:

16 Comparison of NNLO PDF luminosity functions NNLO trends are similar to those observed at NLO

17 Comparison of NNLO predictions

18 PDF4LHC recommendations

19 PDF4LHC recommendations(arxiv: ) Of course, there is the freedom/encouragement to use any individual PDF desired for comparison to measured cross sections. This has been the norm for the 2010 LHC results.

20 Update: CT10 and CT12 CT10 NNLO complements CT10 NLO PDF set only pre-lhc CT10 data same input parameters and parametrization forms as in CT10 NLO PDFs NLO and NNLO PDFs produce about same χ 2 shapes of CT10 NNLO PDFs have evolved compared to CT10 NLO as a result of order α s 2 contributions, updated EWK contributions and revised statistical procedures CT12 NLO and NNLO include LHC W and Z rapidity data include ATLAS and CMS jet data include HERA 2011 F L data α s (mz) fixed at for both NLO and NNLO fits in accordance with world average

21 CT10W NNLO compared to MSTW08 NNLO

22 CT10W NNLO vs CT10W NLO (preliminary) NNLO typically within NLO error bands figures from Marco Guzzi NNLO error bands are of similar size (determined primarily by experimental errors)

23 Some NNLO comparisons CT10 NLO used as reference figure from Zhao Li

24 PDF correlations Consider a cross section X(a), a function of the Hessian eigenvectors i th component of gradient of X is Now take 2 cross sections X and Y or one or both can be pdf s Consider the projection of gradients of X and Y onto a circle of radius 1 in the plane of the gradients in the parton parameter space The circle maps onto an ellipse in the XY plane The angle φ between the gradients of X and Y is given by If two cross sections are very correlated, then cosφ~1 uncorrelated, then cosφ~0 anti-correlated, then cosφ~-1 The ellipse itself is given by

25 Define a correlation cosine between two quantities Correlations Z tt If two cross sections are very correlated, then cosφ~1 uncorrelated, then cosφ~0 anti-correlated, then cosφ~-1

26 from PDF4LHC report (CTEQ6.6) so if two cross sections X and Y are correlated, then PDF uncertainty on ratio is reduced, on sum or differenence is increased

27 Used for LHC Higgs searches

28 in YR2 report

29 Correlations For the 2 nd Higgs Yellow Report, we expanded the correlation information to include all PDF groups Note that the CTEQ, MSTW, NNPDF correlations tend to be very similar; correlations for other PDFs (like EPA mileage) can vary

30 Correlations For the 2 nd Higgs Yellow Report, we expanded the correlation information to include all PDF groups Note that the CTEQ, MSTW, NNPDF correlations tend to be very similar; correlations for other PDFs (like EPA mileage) can vary

31 Background correlations

32 Correlations, strangeness and ATLAS W/Z cross sections There s a strong correlation for the ratio of the W to Z cross section at the LHC to the strange quark distribution Nadolsky et al, PRD78 (2008) ATLAS analysis of W and Z production (taking correlated errors into account) suggests that s (x,q) d (x,q) = at x=0.023 and Q 2 =1.9 GeV 2 Consistent within errors of CT10 more exploration of strange uncertainty in CT12

33 More benchmarking 2 studies in 2011 Les Houches proceedings( ) Benchmarking for inclusive DIS cross sections with S. Alekhin, A. Glazov, A. Guffanti, P. Nadolsky, and J. Rojo excellent agreement observed between CTEQ code with alternative DIS calculation provided by A. Guffanti Benchmark comparison of NLO jet cross sections J. Gao, Z. Liang, H.-L. Lai, P. Nadolsky, D. Soper, C.-P. Yuan compare EKS results with FastNLO (NLOJET++) excellent agreement between the two if care is taken on settings for jet algorithm, recombination scheme, QCD scale choices

34 Followup for PDF4LHC Study of NNLO PDFs from all 6 PDF groups drawing from what Graeme has done, but now including CT10/12 NNLO detailed comparisons to LHC data which have provided detailed correlated systematic error information, keeping track of required systematic error shifts, normalizations, etc ATLAS W/Z rapidity distributions ATLAS inclusive jet cross section data leading to a new recommendation

35 From 7 to 8 TeV little change in uncertainty, as you might expect

36 and finally percentage of articles to the archive that mention the following phrases so if we re willing to wait a few years, we ll beat out supersymmetry; depending on what happens in 2012, maybe even Higgs

37 a continuation of the series that started in Trento (2010) and St. Andrews (2011)

38

39 LHC: W, Z cross sections ATLAS-CONF Many of the experimental/theory errors cancel with the ratio CT10 NNLO prediction Of course, there is much additional information that will be used in PDF fits, such as the Z rapidity distribution and the W asymmetry. ATLAS W/Z cross section ratio in good agreement with NNLO predictions from the PDF groups shown PDF4LHC CMS PAS EWK

40 LHC: W/Z ratios CT10 NNLO Total W/Z ratio from ATLAS in good agreement with theory, but separate W + /Z and W - /Z ratios show some differences (at 1 sigma level) for some of PDFs CMS results for W,Z use PDF4LHC recipe for NNLO; good agreement with theory

41 ATLAS inclusive jets Important to carry predictions out over wide rapidity range. New physics tends to be central. Old physics (PDFs) has an impact on all rapidity regions. This data (or higher statistics version can be fed back into global PDF fits and can/will have impact, especially on high x gluon. ATLAS-CONF but the use in global PDF fits is possible only once detailed correlated systematic error information is made available. For jets, systematic errors are much more important than statistical errors. This has been done for the ATLAS (but not CMS) 2010 data.

42 ATLAS: inclusive jets Important to use more than one jet size. Different dependence on underlying event, fragmentation and also on perturbative prediction. The data tend to fall off faster at high p T /high y. Is this due just to PDFs or is there other physics also involved?

43 CT12 NLO predictions for ATLAS jet production (preliminary) CT12 NLO PDFs predict smaller jet cross sections at large p T than CT10 χ 2 /dof=0.72 (0.98) for CT12 NLO (CT10 NLO) figures from Jun Gao

44 Now to scale dependence Write cross section indicating explicit scale-dependent terms First term (lowest order) in (3) leads to monotonically decreasing behavior as scale increases (the LO piece) Second term is negative for µ<p T, positive for µ>p T Third term is negative for factorization scale M < p T Fourth term has same dependence as lowest order term Thus, lines one and four give contributions which decrease monotonically with increasing scale while lines two and three start out negative, reach zero when the scales are equal to p T, and are positive for larger scales At NLO, result is a roughly parabolic behavior (if you re lucky) Note that each of these terms depends on the kinematics of the cross section under investigation (1) (2) (3) (4) from CHS

45 It s useful to use a log-log scale since perturbative QCD is logarithmic Note that there s a saddle region, and a saddle point, where locally there is no slope for the cross section with respect to the two scales This is kind of the golden point and typically around the expected scale (p T jet in this case)

46 Scale choices Take inclusive jet production at the LHC Canonical scale choice at the LHC is µ r =µ f =1.0*p T CDF used 0.5p T CTEQ6.6/CT10 used this scale for determination of PDFs R=0.4 antikt new CT PDFs use p T Close to saddle point for low p T But saddle point moves down for higher p T (and the saddle region rotates) Our typical scale choices don t work for all LHC kinematics; more extreme movements for some of measured cross sections Rather than look for some magic formula, we should try to understand what is going on the kinematic/scale point-of-view

47 Scale dependence also depends on jet size R=0.4 antikt R=0.6 antikt

48 Scale dependence depends on rapidity The saddle point tends to move upwards in scale as the rapidity increases Is the physics changing; no, just the kinematics

49 Look for saddle point position Position of saddle point Black circles Red squares Green triangles Blue triangles Magenta crosses using a Python script written by Jessie Muir, (now) a Cambridge (Part 3) student

50 µ R increases with y*/y max y*=(y j1 -y j2 )/2 Black circles Red squares Green triangles Blue triangles Magenta crosses

51 µ F increases with y*/y max Black circles Red squares Green triangles Blue triangles Magenta crosses Note: maybe no true saddle points at high y* and high mass, so script has trouble finding them; there are still flat places